A supervised sparsity-based wavelet feature for bearing fault diagnosis

摘要: This paper proposes a supervised sparsity-based wavelet feature (SSWF) for the detection of bearing fault, which combines packet transform (WPT) and sparse coding. SSWF is extracted from vibration signals by four main steps: (1) construct WPT vector using fault-related coefficients; (2) design structured dictionary that signal characteristics class information; (3) use to implement coding vectors, can be solved basis pursuit (BP) (4) calculate coefficients. During process, detect fault occurrence signal. Sparse based on find robust representation at same time, integrate information. Therefore, able stably discriminatively reflect different types, indicates its potential in diagnosis. Experiments two cases are conducted verify advantages faults.